Ben,

That's quite a project, and very interesting. You should know that there's a longstanding cry for Archinect.com to standardize the salary survey questionnaire for just this purpose.

Thank you for posting.

Wow!

ok...who are these 23 year olds making 320k? these sound suspicious since there is no one making that much at an OLDER age. besides that, thanks for the info, bregnier! interesting to see.....


it is too bad there is not a 0 list for the unemployed, though.

this is impressive - I'd be curious how this starts to break down by region and/or by year of report.

Yeah... only noticed those six-figure twenty year olds after I'd made the plot. A bit doubtful to say the least. I tried to remove all of the "0" answers as it's not really a true survey and thus proportions are off. What I really wanted to do was show disparities between men and women but the ManyEyes scatter plot tool doesn't differentiate between data sets with color or shape.

toasteroven, I was thinking of doing some sort of heatmap for regions, but the location section is incredibly messy -- it took me hours just to filter out the US entries (didn't want to deal with currency conversion or cost of living adjustment).

The next iteration of the survey definitely needs to have some in-form data control. For instance, the fact that the salary entry is just a text field is ridiculous... if someone enters 35 does this mean $35.000 or $35/hour?

I'd estimate it would take about 2-3 days to clean up all of the data to get all of it up to spec and delete the uncertain or bullshit entries. If anyone wants to take that on be my guest. (Google Refine is an awesome tool if you're interested.)

this is awesome - great job!

by my (very rough) calculations, the line of best fit is:
2.75 x age - 30 = salary in thousands

for example:
2.75 x (25 years old) - 30 = $38,750
2.75 x (30 years old) - 30 = $52,500
2.75 x (50 years old) - 30 = $107,500


and yes, who are these 23-25 year olds making $300,000+?

Ignore the outliers, concentrate on evaluating the valuable data.

While that trend line looks pretty accurate, it's also important to note the wide spread of the data, regardless of age. Anybody want to calculate the standard deviation?

Wow - that's a lot of work and makes the salary poll data, even if not perfect, EVER so much more usable!

Do we need to mass-email your teacher and tell him/her to give you an A? (just kidding)

One factor that jumps out at me (after looking at the spreadsheet) is how many people chose to participate in salary poll pre- and post-recession. Unwillingness to share bad news about yourself is telling.

Which brings us to the biggest problem of charting this data: the recession has broken any meaningful patterns that could have been extracted from this model.

Back when I was 23, I was making $320k, now I hardly make half of that. I don't need 'science' to tell me things kind of suck.

A very interesting exercise, although there's clearly some "issues" with the accuracy of the data ...

For example, if one looks at some of the more suspicious outliers in the CSV raw data and compares the pay figures with the 'explanation' field, one finds the following examples:

$418,000 - Unlicensed. 10 years experience. No Health. 1% Matching IRA. Underpaid

$350,000 - B.ARCH. 1 yr. exp.. health. no paid OT

$340,000 - MArch No Experience

$320,000 - Intern no benefits 2 years exp

Looks to me like these sorts of entires have a missing decimal about one digit from the right -- then, they might appear more reasonable.

Nevertheless, a very useful way to view data ... thanks for doing this work.

not to threadjack, but what software are you guys using for creating vector/illustrator compatible graphs from raw data?

IMO if you're looking for power, flexibility, and interactivity, you can't beat the Processing language (processing.org). Protovis for java is also good. If you happen to have a license of mathematica, that's a graphics powerhouse that doesn't require as much programming (although still needs a lot of math).

The ManyEyes tool above is also good for quick visualization, but lacks flexibility.

quizzical, I think you're right that those people just misplaced their comma.

Or they might have been paid in pesos. Either way, I need to see a pie chart of this information to pass judgement. I find charts confusing.

I was about to say.

How do we know these people aren't foreigners who use a dollar system but not necessarily the American dollar?

$10,000 USD is equal to ~$80,000 HKD (Hong Kong).

So, $50,000 USD entry-level position in Hong Kong would pay roughly $500,000 HKD a year.

Oh the sweet, sweet ironing.

Someone deleted that post?!?!?

To strix:
The scatterplot data is US respondents only. Some of the international entries do have unclear currency designation. It is possible that some of the outports could even be my fault as the result of my methods for cleaning the data-- I tried to be careful but I am no expert at this.

That should say outLIERs. Missed the autocorrect. Should never make forum posts whilst teaching my screaming twin two year olds the value of sharing.

Whoa. Whoever deleted Stixies embarrassing post, bring it back. We may be illiterate but we're still #1.

USA! USA!